Claude Code: The Terminal AI

Let me be honest with you: I was skeptical at first. I’ve been using Claude.ai on the web for a while now, and it works great. But every time I needed help with code, I was doing this dance switch tabs, paste the code, read the answer, switch back, type it in. It gets old fast.

Then I tried Claude Code, the CLI version. And I genuinely didn’t expect to like it as much as I did.

This article will walk you through what Claude Code actually is, how it’s fundamentally different from the web app or desktop app, and how to get it installed and running on your machine. I’ll also hand you a set of commands you can start using right away no guessing, no Googling.

So What Exactly Is Claude Code?

Claude Code is Anthropic’s terminal-based AI coding agent. You install it like any other CLI tool, run ‘claude’ in your terminal, and it opens an interactive session right there no browser, no separate app, no tab-switching.

But it’s not just a chatbot that happens to live in your terminal. That’s the important part to understand. Claude Code can:

• Read files in your project directory

• Write and edit files directly on disk

• Run shell commands on your behalf

• Manage Git operations like commits, diffs, branch switches

• Run your tests and fix what breaks

• Connect to external services via MCP (Model Context Protocol)

It understands your project as a whole, not just a pasted snippet. That’s a fundamentally different thing from having a chat window open next to your editor.

Claude Code vs The Web App vs The Desktop App

A lot of people don’t realize these are three actually different products. Here’s a quick breakdown:

The web app is great for questions and exploration. The desktop app is handy for quick access over any window. But when you actually want to build something when you have a real project with real files Claude Code is the one designed for that job.

Why Developers Are Switching to the CLI Version

There are a few things about working in the terminal that just click differently once you start using Claude Code.

It Lives Where Your Code Lives

When you run ‘claude’ inside your project folder, it knows the project. It can read your files, check your directory structure, and understand context that you’d have to manually copy-paste into a web chat. That alone saves minutes every session.

It Can Actually Execute Things

Not just suggesting things but they do them. You can ask it to run your test suite, and it will. Ask it to create a new feature branch, push a commit, or refactor a module. It’ll ask for your permission before doing anything destructive, but it can carry out multi-step tasks end to end.

It Fits into Your Workflow

If you’re already working in VS Code, Vim, or any editor that has a terminal panel, Claude Code fits right in. No separate window, no context switching. Some developers keep it running in a terminal split alongside their editor all day.

Automation and Scripting

Claude Code has a non-interactive “print” mode (claude -p "your prompt") that’s designed for scripting and CI/CD pipelines. You can run AI-powered code reviews in GitHub Actions, automate test fixing, or integrate it into build scripts. That’s not possible with any of the GUI versions.

Before You Install: Pre-requisite

• Operating System: macOS 13+, Windows 10/11, Ubuntu 20.04+, Debian 10+, or Alpine Linux 3.19+

• Hardware: 4 GB RAM minimum, x64 or ARM64 processor

• Internet connection: required for auth and AI calls

• Anthropic account: Pro ($20/month), Max, Team, or Enterprise — the free plan does not include Claude Code

• For Windows: Git for Windows is optional but recommended (gives you the Bash tool)

Tip: You don’t need Node.js for the native installer (recommended). It’s self-contained. Node.js is only needed if you choose the npm install path.

Installation: Step by Step

The easiest and most reliable way to install Claude Code is the native installer. It handles everything with no dependencies to manage, and it auto-updates in the background so you always have the latest version.

Step 1 — Open Your Terminal

On macOS, open Terminal (or iTerm2, Warp — whatever you use). On Linux, open your preferred terminal emulator. On Windows, open PowerShell (search for it in Start).

Step 2 — Run the Install Command

Copy and paste the command for your platform:

macOS / Linux / WSL: curl -fsSL https://claude.ai/install.sh | bash

Windows PowerShell: irm https://claude.ai/install.ps1 | iex

Windows CMD (Command Prompt): curl -fsSL https://claude.ai/install.cmd -o install.cmd && install.cmd && del install.cmd

The installer downloads the binary, places it in ~/.local/bin/claude (or %USERPROFILE%\.local\bin\claude.exe on Windows), and adds it to your PATH automatically.

Step 3 — Open a New Terminal Window

This is important and don’t skip it. After installation, your current terminal session doesn’t know about the new PATH entry yet. Close the window and open a fresh one, or run:

# For bash users: source ~/.bashrc

# For zsh users (default on macOS): source ~/.zshrc

Step 4 — Verify the Installation

Run these two commands to confirm everything is in order:

# Check version

claude --version

# Run full diagnostics

claude doctor

claude --version should print a version number. claude doctor runs a full environment check and tells you if anything is misconfigured. It’s a good habit to run it after installation.

Step 5 — Authenticate

The first time you run claude, it needs to know who you are. Just run it: claude

Claude Code will automatically open your default browser to the Anthropic login page. Sign in with your Anthropic account (the same one tied to your Pro, Max, Team, or Enterprise subscription). Once you authorize the connection in the browser, your terminal session picks it up automatically and no copy-pasting tokens.

Tip: Authentication is a one-time thing. After the first login, you don’t need to re-authenticate. Claude Code stores your session locally.

Your First Session: Let’s Actually Use It

With everything installed and authenticated, navigate to any project folder you’re working on and launch Claude Code:

cd ~/my-project

claude

You’ll land in an interactive terminal session. It looks a bit like a chat prompt, but it has access to your entire project directory. From here you can just talk to it like you would in the web app except now it can actually reach into your files.

Here are some things to try in your first session:

• what files are in this project? — it reads your directory and describes it

• explain what this project does — it reads your code and gives you a summary

• Ask it to fix a bug: the login function in auth.js is throwing an error, can you fix it?

• Ask it to write tests: write unit tests for the UserService class

Essential Commands Every New User Should Know

Claude Code has two layers of commands: the CLI flags you use when launching it from your terminal, and the slash commands you type inside an active session. Here’s what you’ll actually use day to day.

CLI Commands (Run from Terminal)

Slash Commands (Use Inside an Active Session)

Once you’re inside Claude Code (after running claude), these commands control the session:

Special Syntax Inside a Session

Two quick tricks that save a lot of time:

• Reference a file with @: Type @./src/App.tsx to point Claude at a specific file. Example: review @./src/auth.js for security issues

• Run a shell command with !: Prefix any shell command with ! to run it without leaving the session. Example: !npm test

A Real-World Example

Let’s say you’ve got a Node.js project and you want Claude Code to write tests for a function that doesn’t have any. Here’s roughly how that session looks:

That whole flow from “write tests” to “tests passing” without leaving the terminal or touching a browser tab.

Pricing

Claude Code is not available on the free Claude.ai plan. You need a paid subscription:

• Claude Pro ($20/month): Works for most individual developers. Includes Claude Code in the terminal and Claude on the web.

• Claude Max ($100–$200/month): Higher rate limits for heavy usage. If you’re running long sessions or lots of automation, this is where you end up.

• Team / Enterprise: Adds team management, SSO, admin controls, and higher organizational limits.

• Anthropic API (Console): Pay-per-token access. Works with Claude Code if you prefer that over a flat subscription.

Wrapping Up

Getting Claude Code set up genuinely takes about five minutes — the install command is a one-liner, authentication is automatic, and you’re in. What surprised me wasn’t the installation; it was how quickly it becomes something you just keep open.

The terminal version isn’t for everyone. If you mostly use Claude for writing, summarizing, or casual questions, the web app works perfectly well. But if you’re a developer who spends most of your day in a terminal, and you want an AI that can actually touch your files and run your code — Claude Code is the version built for you.

Start by running it in a real project, not a toy example. Ask it to explain something complicated in your codebase or write a test for a function that doesn’t have one. That’s where the difference between a chatbot and an actual coding agent becomes obvious.

If you get stuck anywhere in the process, feel free to drop a question in the comments. Happy to help.